1

ROLLER BEARING FAILURES IN REDUCTION GEAR CAUSED BY INADEQUATE DAMPING BY ELASTIC COUPLINGS FOR LOW ORDER EXCITATIONS

~by Herbert Roeser, Trans Marine Propulsion Systems, Inc. Seattle

Flexible couplings provide a convenient means for modifying and improving the natural frequency characteristics of a system, often having significant damping effect under resonant conditions. The ability to reduce resonant vibration amplitudes is usually confined to modes of vibration having dominant relative angular deflections across the coupling elements, since the actual damping torque is proportional to the rate of change of angular displacement in the coupling. This means that the frequency of the vibration has a large effect on the damping torque in an elastic coupling. This is the reason why the lower order vibrations are more difficult to dampen out. The power loss in the coupling is related to the amount of damping provided by the coupling. Due to this, it is often necessary to tune the system such that there is no resonant condition within the operating range of the system that can be excited by the low engine orders. If this is not accomplished, low order vibration amplitudes can be transmitted to the reduction gears by the coupling due to insufficient damping at resonance condition.

1. Introduction:

The problem an engineer faces today when commencing a torsional vibration analysis of a diesel engine propulsion plant is the ever increasing complexity of these plants. One aspect of the greater complexity is the absolute increase in the calculations that are required due to the use of PTI and PTO gears with attached generators and multiple geared diesel engines. With respect to the diesel engine development, in order to achieve a better fuel economy, the trend is lower running speeds; longer stroke-to-bore ratios and higher combustion pressures.

In order to predict the behavior of any physical

system, a model suitable for mathematical analysis is of great importance. The requirement of the model is to be able to predict the behavior of the system with sufficient accuracy. In the early days of torsional vibration analysis, a sufficient accurate model meant, a model which is able to predict the first few modes of vibration and the vibratory stress amplitudes in the shaft system at the calculated critical speeds. This type of simple analysis is no longer sufficient. The ever ongoing development of higher output diesel engines and

increased complexity in operation and design of propulsion plants have proven the need for mathematical models capable of predicting behavior at non-resonant conditions. This includes the effects of damping, vibratory torques in reduction gears and elastic couplings, and heat losses in the elastic couplings and torsional dampers. The calculations must involve all expected operating parameters of the analyzed plant. In this paper, we shall describe a case where the damping in the elastic coupling was not sufficient to dampen the low frequency vibration which eventually led to early roller bearing failure in the gears in question. The intension here is not to describe the method of mathematical modeling, but to use simple mathematical models to better explain the problem described in this paper.

2. Equation of Motion:

Mass-elastic properties of the components in a propulsion system are often specified by each manufacture. In addition to the stiffness and inertia values, the shaft and mass damping are part of a complete mass-elastic system. The total mass-elastic system of a propulsion plant constitutes the

2

mathematical model to be analyzed. By applying Newton’s second law for a rotating body, the so-called equation of motion is derived. If the general case with damping and an external torque is considered, the equation of motion for a mass becomes,

The first term in this equation is the inertia

torque; the second and third terms are the shaft (relative) damping torque and the mass (absolute) damping torque, respectively. The fourth term is the spring or stiffness torque and the last term is the external or excitation torque.

A propulsion system consists of several masses

connected by shafts and the equation of motion is applied to each mass in the system. The result is a set of linear differential equations where the number of equations is equal to the number of masses in the system. The free modes of vibration and the corresponding natural frequencies are found by setting the damping torque and external torque equal to zero and solving the resulting eigen value problem by Holzer tabulations or a root finding technique. 3. Method of Analysis:

A torsional vibration analysis consists of two parts. The first part is natural frequencies and mode shape calculations, and the second part is the forced-damped calculations. After the mass-elastic system has been established, the analysis is in general performed in the following manner. The equation of motion is setup for each mass in the system. The result is sets of linear differential equations that are transformed into a characteristic system of linear algebraic equations. These are solved as an eigen value problem for the natural frequency calculations, and as a system of simultaneous equations for the forced damped calculations.

The natural frequency and mode shape calculations will determine the resonant or critical speeds and the relative deflection curve of each mode of vibration. The results should be presented in such a way that the resonant speed at each mode of vibration is identified, including the vector summation of each order. The vector sum is calculated based on the firing order and deflection curve and indicates the relative significance of each order for a specific mode of vibration.

The conventional Holzer tabulation for

undamped free vibrations with the natural frequencies, relative deflections, inertia torques, and phase vector summations, still serves as a way of obtaining an overall view of vibration characteristics of the propulsion system, and forms the basis for the detailed forced-damped vibration analysis. 4. Forced-Damped Calculations:

In order to determine the response of a complex propulsion system with several sources of damping, and to obtain a more realistic picture of the non-resonant conditions, a forced damped analysis is necessary. The system of equations is solved for each order by using the harmonic components of the tangential efforts as the external torque, including the propeller excitation. The calculated results for each order and mode shape are added as a synthesis, taking into account the phase relations in order to find the overall vibration amplitudes or power dissipation in the analyzed propulsion system.

In general, torsional vibration calculations are

performed assuming there is equal excitation from each cylinder in the engine. However, based on experience and measurements it has been shown that the lower order harmonics are often more significant than calculated assuming equal excitation from each cylinder. This can be explained by the fact that the excitation from each cylinder is not equal or uniform, but shows a lot of

3

variations due to changes in firing pressure and ignition characteristics. This irregularity will excite the low order vibration and must be considered as part of a torsional vibration analysis.

At this point it should be noted that it is very

difficult to evaluate the resulting engine excitation torque, since the excitations will depend on the rate and magnitude of the irregularity of each cylinder and on the overall distribution of impulses throughout the engine. Due to this, in some cases, based on experience, it may be sufficient to consider two excitation modes; the first is a “normal” combustion mode with equal excitation from each cylinder, the other is a “misfiring” mode where one cylinder is operating without combustion (compression only). The misfiring condition is considered a failure mode, but in engines with large numbers of cylinders, this condition will simulate the normal stochastic irregularity reasonably well. On the other hand, for engines with less than 12 cylinders, the misfiring condition may be too unfavorable to simulate the irregularity and the actual vibratory response of the system is between the normal combustion mode (equal excitation) and the misfiring condition. 5. Example: Premature Failure of Roller

Bearing in PTI Gears:

The propulsion plants in question consist of a large seven cylinder two-stroke engine, rated 20588 kW at 102 rpm. The five bladed controllable pitch propeller is directly driven by the main engine. In addition, an asynchronous electrical motor capable of delivering 3822 kW at 1200 rpm via a PTI gear through an elastic coupling to the propeller increases the total output to 24410 kW. Figure 1 shows the propulsion system arrangement:

Figure 1 Arrangement of Propulsion Plant

The propulsion plant arrangement shown in

Figure 1 is similar to those installed on three (3) large container vessels in order for them to reach a service speed of 22 knots. After approximately one year of service all three vessels encountered PTI gear failures. During the investigation several different assumptions regarding the failures were discussed, due to the fact that in two vessels the failures related to roller bearings and on the third vessel it related to a gear tooth failure.

During our investigation, it became quite clear

that the problem with these failures where directly related to the dynamic loading of the PTI gears in service. This dynamic loading must relate to the torsional behavior of the system that affected the roller bearing service life in a very negative way. Apart from the service restrictions specified by the engine manufacturer, based on their torsional vibration analysis, we had to commence our own torsional vibration analysis. In particular, we were interested to evaluate the vibratory torque amplitudes transmitted through the PTI gear, since these torque amplitudes has a large influence on the dynamic loading of the roller bearings and gear teeth.

Based on our own calculations and in

comparison with the engine manufactures calculated results (steady state calculations) we could not detect any parameters that should cause

4

any concerns if the service restrictions specified by the engine manufacturer were followed. It was then decided that actual measurement should be commenced on one of the vessels which at that time had been repaired.

5.1 Measurement Setup:

Torsional vibrations were measured with a Brǘel & Kjaer torsional vibration meter type 2523, which is a laser system for non-contact measurements of angular vibration velocity or angular vibration displacement.

Angular vibration displacements up to 17.2

degrees can be measured in the frequency range of 0.3 Hz to 1 KHz. The signal from the laser transducer is fed into a Brűel & Kjaer 2-channel real time analyzer type 2145. A tachometer is used to measure the engine speed and the output signal is used by the real time analyzer for order tracking. The measurements were recorded on tape for further analysis.

5.2 Verification of Theoretical Calculations:

In a case like this, it becomes important to verify the mathematical model used in the theoretical calculations, particularly since the natural frequencies and mode shape calculations do not include any damping in the system (Free vibration).

Based on our own experience the best way to

verify the calculated natural frequencies and mode shape calculations is to measure the angular displacement and acceleration at the free end of the crankshaft.

Fig 2 Average Vibratory Amplitudes

The graph in Figure 2 shows a very good

correlation between the calculated and measured results. The graphs include the synthesis and the first seven engine orders of vibratory angular amplitudes. The measured values are average values, measured over a time interval of two minutes. At the time of these measurements the main engine load was 80% of MCR, and the booster load was 100%. As can be seen, the measured amplitudes are higher than the theoretical calculated amplitudes for normal firing (equal excitation from each cylinder). This is expected due to the normal stochastic irregularities between the cylinders as explained earlier in this paper. Additionally, it is important to notice that the larger the stochastic irregularities between the cylinders, the more significant is the influence from the engine orders, in this case, 1st, 2nd, 3rd and 4th.

5.3 Torsional Vibration Measurement on the PTI Gears:

These measurements were very important in

this case, and had to be planned carefully. Since we were concerned about the lower engine orders and their influence on the PTI gear, it was determined to measure the lower engine orders

5

across the gear tooth coupling between the asynchronous electrical motor and the high-speed pinion gear. Figure 3 shows a picture of the measurement setup.

Fig 3 Vibration Measurement Setup

Figures 4 and 5 show the measured vibratory amplitudes at the booster motor and gear side respectively.

Fig 4 Measured Angular Displacements

on the Booster Motor Side

Fig 5 Measured Angular Displacements on the

Gear Side

As can be seen from the measured results, the vibratory amplitudes across the gear tooth coupling are independent of booster motor load. The engine order that gives the highest contribution is the 1st, 2nd, and 3rd engine orders. In general, the amplitudes are somewhat higher on the gear side than on the motor side, this could be related to the damping in the gear tooth coupling, even though this parameter is not important. In this case, the important parameters are the vibratory amplitudes on the PTI gear side.

6. Roller Bearing Life: Bearing life is generally defined in hours of

operation before the bearing fails under given loading conditions. It is very difficult to predict the life of an individual bearing under given loading conditions because it depends on several factors like lubrication, loading and speed of the bearing. In practice, apparently similar bearings exhibit different bearing life under exactly similar loading conditions. Hence, bearing life is described statistically in terms of life. life of a bearing is the number of hours, 90% of exactly similar bearings run under the given load conditions before failing. 10% of the bearings will

6

fail before the life is reached. The most common way to calculate life of a bearing is by using the formula given by ISO.

Where

= life of bearing in hours C = dynamic load rating (load at which at least 90% of identical bearings complete one million revolutions) P = equivalent dynamic bearing load p = exponent = (3 for ball bearing; 10/3 for roller bearing) RPM = revolutions per minute

Based on our findings, particularly by our

measured data, it became important to determine how large the variation in vibratory amplitudes was over a certain time interval for the lower engine orders. As can be seen from Figure 6, the lower engine orders show a relative large variation in vibratory amplitudes over the measured time intervals, particularly the 1st engine order.

Fig 6 Time Capture of Measurement

As we attempted to evaluate the actual dynamic loading of the roller bearings involved, accounting for the large variations in vibratory torque amplitudes, particularly from the 1st engine order became necessary.

Roller bearing life calculations were

commenced for two different loading conditions, case #1 and case #2.

#1. Based on mean transmitted torque at 3822 kW@ 1200rpm. #2. Based on mean transmitted torque at 3822kW@1200rpm and in addition, the contribution from the measured vibratory torques transmitted through the gears.

Based on the calculated results and as expected,

the roller bearing’s B10 life is drastically reduced when the dynamic loads due to low order engine excitations are included. From the calculated results, it can clearly be seen that all roller bearings are very highly loaded, and as a result, reliability and service life are severely compromised.

7. Dynamic Roller Bearing Life Calculations:

7

Gear Data:

= 27; = 68; = 26; = 121

= 2.5185; = 4.6538

= 11.7206

Approximate sizes from drawing, scale 1:5

= 11.6564

Gear Forces: Assuming Helix angle = = 12.5° and Pressure angle = = 20.0°

= 0.3728

Gear 1 – Gear 2:

= 30416.75 N-m

= 193122.2222 N

= 70296.4889 N

= 42815.1967 N

Gear 3 – Gear 4:

= 76283.2778 N-m

= 346742.1718 N

= 126214.1505 N

8

= 76872.7395 N

= 354543.8707 N-m Bearing Reactions: Defined as the force exerted by the bearing on the shaft

Adding moments in i direction

= 0 = 0

= 0 N

= 0 = 0

= 0

N Adding moments in j direction

= 0 = 0

= 0 N

= 0 = 0

= 0 N

= 93024.0919 N

= 112494.2505 N

Adding moments in i direction

= 0

N

= 0

N

9

Adding moments in j direction

= 0 = 0

= 0

N

= 0 = 0

= 0

N

= 280677.5468 N

= 265458.3277 N

Adding moments in i direction

= 0

N

= 0 N Adding moments in j direction

= 0 = 0 N

= 0 = 0 N

= 227076.2105 N

= 141922.6316 N Hertzian Stress / Service Life of Roller Bearings:

where, c = Dynamic basic load N

P = Load = Lift time in million revolutions of the

bearing With a static inner ring, the number of load cycles on the inner ring per revolution are:

10

z = number of rollers

= Diameter of roller = Mean diameter

= number of load cycles on the inner ring until

10% fail (in million cycles)

= Number of active rollers

= Maximum load on roller

where, = Diameter of the inner race E = Modulus of elasticity Calculation of SKF bearing N216 Hertzian stress, which gives 10% failures after one million loads cycles on the inner ring 1) = 100 2) Number of load cycles on the inner ring /

revolution

= 10.2239 cycles / revolution 3) Number of bearing revolution (in million

cycles)

= 9.7810

4)

for N216 – N (from SKF bearing catalogue)

= 52.9698

5) Number of active rollers

= 3.6

6) Maximum load on roller

N

7) Hertzian stress

= 2432.9862 = 352874.7226 psi Dynamic Bearing Life Calculations:

Fig 7 PTI Gear Bearing Arrangement

11

Bearing #103: A & B Cylindrical Roller bearing: NV 2052 E MA C3

Load Rating Dynamic Static

267000 lbs 467000 lbs

1190 kN 2080 kN

based on mean transmitted torque: Highest radial load at support B: = 113 kN

= 35554.61 hours

based on mean transmitted torque and torsional vibratory torque:

Vibratory torque = = 15 kN-m Mean transmitted torque = = 30.4 kN-m Radial Load:

= 55.7566 kN

= 134.3371 kN

= 164.5124 kN

= 91.6629 kN

= 61.4876 kN

+

P = 127.0164 kN

hours Bearing #104: C Cylindrical Roller bearing: NV 2348 E MA C3

Load Rating Dynamic Static

465000 lbs 710000 lbs 2070 kN 3160 kN

based on mean transmitted torque: Highest radial load at support B: = 281 kN

= 27207.5568 hours

based on mean transmitted torque and torsional vibratory torque:

12

Vibratory torque = = 70 kN-m Mean transmitted torque = = 76.3 kN-m Radial Load:

= 257.80 kN

= 379.6558 kN

= 519.1761 kN

= 182.3442 kN

= 42.8239 kN

+

P = 377.5324 kN

hours Bearing #105: D Spherical Roller bearing: 22348 MB W 33

Load Rating

Dynamic Static

6x lbs 0.35 2 2.9 88x lbs 1.9 2671 kN 3917 kN

X = 1 for spherical roller bearing

= 34057.5428 N = 34.10 kN

265458.3277 N = 265.50 kN

= 0.13 e = 0.35

based on mean transmitted torque:

=

= 35881.3188 hours

13

based on mean transmitted torque and torsional vibratory torque:

= 265.5 kN

= 34.10 kN Vibratory torque = = 70 kN-m Mean transmitted torque = = 76.3 kN-m Radial Load:

= 342.5780 kN

Axial Load:

= 31.2844 kN

Radial Load:

= 358.7217 kN = 490.5570 kN

= 172.2783 kN

= 40.4429 kN

+

P = 365.7189 kN Axial Load:

= 46.0780 kN

= 63.0174 kN

= 22.1220 kN

= 5.1827 kN

+

P = 45.8231 kN

= 12546.2040 hours

Bearing #106: Next to A Angular contact ball bearing: QJ 1052 N2 MPA C3

Load Rating Dynamic Static

120000 lbs 225000 lbs

540 kN 1000 kN The angular contact ball bearing is designed to carry only axial load.

based on mean transmitted torque:

= 42815 N = 42.80 kN

= 27894.4290 hours

Even though this bearing was affected by the vibratory torque, the reduction in bearing life was very small. This bearing did not fail in any of the three applications. Only the bearings #104 and #105 failed in all three applications. As can be seen

14

from the table below, bearing #104 and #105 show a significant reduction in the bearing life.

8. Conclusion:

In the case described, the failures are directly related to the damping capacity of the elastic couplings at low frequencies. In this case, particularly the PTI gear manufacturer should have considered this, since the roller bearings service life was adversely affected by these particular parameters. This is also the evidence that with modern plant, in order to achieve reliable service life, the amount of calculations and analysis requirements have increased over time. Simple empirical formula calculations no longer represent the dynamics of these complex systems.

Calculated bearing life in hours Bearing Item# Load

Case#1 Load

Case#2 %

Reduction #103 Cylindrical Roller Bearing

35555 24078 32.28

#104 cylindrical Roller Bearing

27208 10167 62.63

#105 Spherical Roller Bearing

35881 12546 65.03